OMAP3: PM: CPUidle: only scale voltage for deeper C-states
[linux-ginger.git] / net / ipv6 / ndisc.c
blobf74e4e2cdd061433345929399b7a7266b32dacec
1 /*
2 * Neighbour Discovery for IPv6
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Mike Shaver <shaver@ingenia.com>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * Changes:
18 * Pierre Ynard : export userland ND options
19 * through netlink (RDNSS support)
20 * Lars Fenneberg : fixed MTU setting on receipt
21 * of an RA.
22 * Janos Farkas : kmalloc failure checks
23 * Alexey Kuznetsov : state machine reworked
24 * and moved to net/core.
25 * Pekka Savola : RFC2461 validation
26 * YOSHIFUJI Hideaki @USAGI : Verify ND options properly
29 /* Set to 3 to get tracing... */
30 #define ND_DEBUG 1
32 #define ND_PRINTK(fmt, args...) do { if (net_ratelimit()) { printk(fmt, ## args); } } while(0)
33 #define ND_NOPRINTK(x...) do { ; } while(0)
34 #define ND_PRINTK0 ND_PRINTK
35 #define ND_PRINTK1 ND_NOPRINTK
36 #define ND_PRINTK2 ND_NOPRINTK
37 #define ND_PRINTK3 ND_NOPRINTK
38 #if ND_DEBUG >= 1
39 #undef ND_PRINTK1
40 #define ND_PRINTK1 ND_PRINTK
41 #endif
42 #if ND_DEBUG >= 2
43 #undef ND_PRINTK2
44 #define ND_PRINTK2 ND_PRINTK
45 #endif
46 #if ND_DEBUG >= 3
47 #undef ND_PRINTK3
48 #define ND_PRINTK3 ND_PRINTK
49 #endif
51 #include <linux/module.h>
52 #include <linux/errno.h>
53 #include <linux/types.h>
54 #include <linux/socket.h>
55 #include <linux/sockios.h>
56 #include <linux/sched.h>
57 #include <linux/net.h>
58 #include <linux/in6.h>
59 #include <linux/route.h>
60 #include <linux/init.h>
61 #include <linux/rcupdate.h>
62 #ifdef CONFIG_SYSCTL
63 #include <linux/sysctl.h>
64 #endif
66 #include <linux/if_addr.h>
67 #include <linux/if_arp.h>
68 #include <linux/ipv6.h>
69 #include <linux/icmpv6.h>
70 #include <linux/jhash.h>
72 #include <net/sock.h>
73 #include <net/snmp.h>
75 #include <net/ipv6.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
80 #include <net/icmp.h>
82 #include <net/netlink.h>
83 #include <linux/rtnetlink.h>
85 #include <net/flow.h>
86 #include <net/ip6_checksum.h>
87 #include <net/inet_common.h>
88 #include <linux/proc_fs.h>
90 #include <linux/netfilter.h>
91 #include <linux/netfilter_ipv6.h>
93 static u32 ndisc_hash(const void *pkey, const struct net_device *dev);
94 static int ndisc_constructor(struct neighbour *neigh);
95 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
96 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
97 static int pndisc_constructor(struct pneigh_entry *n);
98 static void pndisc_destructor(struct pneigh_entry *n);
99 static void pndisc_redo(struct sk_buff *skb);
101 static const struct neigh_ops ndisc_generic_ops = {
102 .family = AF_INET6,
103 .solicit = ndisc_solicit,
104 .error_report = ndisc_error_report,
105 .output = neigh_resolve_output,
106 .connected_output = neigh_connected_output,
107 .hh_output = dev_queue_xmit,
108 .queue_xmit = dev_queue_xmit,
111 static const struct neigh_ops ndisc_hh_ops = {
112 .family = AF_INET6,
113 .solicit = ndisc_solicit,
114 .error_report = ndisc_error_report,
115 .output = neigh_resolve_output,
116 .connected_output = neigh_resolve_output,
117 .hh_output = dev_queue_xmit,
118 .queue_xmit = dev_queue_xmit,
122 static const struct neigh_ops ndisc_direct_ops = {
123 .family = AF_INET6,
124 .output = dev_queue_xmit,
125 .connected_output = dev_queue_xmit,
126 .hh_output = dev_queue_xmit,
127 .queue_xmit = dev_queue_xmit,
130 struct neigh_table nd_tbl = {
131 .family = AF_INET6,
132 .entry_size = sizeof(struct neighbour) + sizeof(struct in6_addr),
133 .key_len = sizeof(struct in6_addr),
134 .hash = ndisc_hash,
135 .constructor = ndisc_constructor,
136 .pconstructor = pndisc_constructor,
137 .pdestructor = pndisc_destructor,
138 .proxy_redo = pndisc_redo,
139 .id = "ndisc_cache",
140 .parms = {
141 .tbl = &nd_tbl,
142 .base_reachable_time = 30 * HZ,
143 .retrans_time = 1 * HZ,
144 .gc_staletime = 60 * HZ,
145 .reachable_time = 30 * HZ,
146 .delay_probe_time = 5 * HZ,
147 .queue_len = 3,
148 .ucast_probes = 3,
149 .mcast_probes = 3,
150 .anycast_delay = 1 * HZ,
151 .proxy_delay = (8 * HZ) / 10,
152 .proxy_qlen = 64,
154 .gc_interval = 30 * HZ,
155 .gc_thresh1 = 128,
156 .gc_thresh2 = 512,
157 .gc_thresh3 = 1024,
160 /* ND options */
161 struct ndisc_options {
162 struct nd_opt_hdr *nd_opt_array[__ND_OPT_ARRAY_MAX];
163 #ifdef CONFIG_IPV6_ROUTE_INFO
164 struct nd_opt_hdr *nd_opts_ri;
165 struct nd_opt_hdr *nd_opts_ri_end;
166 #endif
167 struct nd_opt_hdr *nd_useropts;
168 struct nd_opt_hdr *nd_useropts_end;
171 #define nd_opts_src_lladdr nd_opt_array[ND_OPT_SOURCE_LL_ADDR]
172 #define nd_opts_tgt_lladdr nd_opt_array[ND_OPT_TARGET_LL_ADDR]
173 #define nd_opts_pi nd_opt_array[ND_OPT_PREFIX_INFO]
174 #define nd_opts_pi_end nd_opt_array[__ND_OPT_PREFIX_INFO_END]
175 #define nd_opts_rh nd_opt_array[ND_OPT_REDIRECT_HDR]
176 #define nd_opts_mtu nd_opt_array[ND_OPT_MTU]
178 #define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
181 * Return the padding between the option length and the start of the
182 * link addr. Currently only IP-over-InfiniBand needs this, although
183 * if RFC 3831 IPv6-over-Fibre Channel is ever implemented it may
184 * also need a pad of 2.
186 static int ndisc_addr_option_pad(unsigned short type)
188 switch (type) {
189 case ARPHRD_INFINIBAND: return 2;
190 default: return 0;
194 static inline int ndisc_opt_addr_space(struct net_device *dev)
196 return NDISC_OPT_SPACE(dev->addr_len + ndisc_addr_option_pad(dev->type));
199 static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
200 unsigned short addr_type)
202 int space = NDISC_OPT_SPACE(data_len);
203 int pad = ndisc_addr_option_pad(addr_type);
205 opt[0] = type;
206 opt[1] = space>>3;
208 memset(opt + 2, 0, pad);
209 opt += pad;
210 space -= pad;
212 memcpy(opt+2, data, data_len);
213 data_len += 2;
214 opt += data_len;
215 if ((space -= data_len) > 0)
216 memset(opt, 0, space);
217 return opt + space;
220 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
221 struct nd_opt_hdr *end)
223 int type;
224 if (!cur || !end || cur >= end)
225 return NULL;
226 type = cur->nd_opt_type;
227 do {
228 cur = ((void *)cur) + (cur->nd_opt_len << 3);
229 } while(cur < end && cur->nd_opt_type != type);
230 return (cur <= end && cur->nd_opt_type == type ? cur : NULL);
233 static inline int ndisc_is_useropt(struct nd_opt_hdr *opt)
235 return (opt->nd_opt_type == ND_OPT_RDNSS);
238 static struct nd_opt_hdr *ndisc_next_useropt(struct nd_opt_hdr *cur,
239 struct nd_opt_hdr *end)
241 if (!cur || !end || cur >= end)
242 return NULL;
243 do {
244 cur = ((void *)cur) + (cur->nd_opt_len << 3);
245 } while(cur < end && !ndisc_is_useropt(cur));
246 return (cur <= end && ndisc_is_useropt(cur) ? cur : NULL);
249 static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
250 struct ndisc_options *ndopts)
252 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
254 if (!nd_opt || opt_len < 0 || !ndopts)
255 return NULL;
256 memset(ndopts, 0, sizeof(*ndopts));
257 while (opt_len) {
258 int l;
259 if (opt_len < sizeof(struct nd_opt_hdr))
260 return NULL;
261 l = nd_opt->nd_opt_len << 3;
262 if (opt_len < l || l == 0)
263 return NULL;
264 switch (nd_opt->nd_opt_type) {
265 case ND_OPT_SOURCE_LL_ADDR:
266 case ND_OPT_TARGET_LL_ADDR:
267 case ND_OPT_MTU:
268 case ND_OPT_REDIRECT_HDR:
269 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
270 ND_PRINTK2(KERN_WARNING
271 "%s(): duplicated ND6 option found: type=%d\n",
272 __func__,
273 nd_opt->nd_opt_type);
274 } else {
275 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
277 break;
278 case ND_OPT_PREFIX_INFO:
279 ndopts->nd_opts_pi_end = nd_opt;
280 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
281 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
282 break;
283 #ifdef CONFIG_IPV6_ROUTE_INFO
284 case ND_OPT_ROUTE_INFO:
285 ndopts->nd_opts_ri_end = nd_opt;
286 if (!ndopts->nd_opts_ri)
287 ndopts->nd_opts_ri = nd_opt;
288 break;
289 #endif
290 default:
291 if (ndisc_is_useropt(nd_opt)) {
292 ndopts->nd_useropts_end = nd_opt;
293 if (!ndopts->nd_useropts)
294 ndopts->nd_useropts = nd_opt;
295 } else {
297 * Unknown options must be silently ignored,
298 * to accommodate future extension to the
299 * protocol.
301 ND_PRINTK2(KERN_NOTICE
302 "%s(): ignored unsupported option; type=%d, len=%d\n",
303 __func__,
304 nd_opt->nd_opt_type, nd_opt->nd_opt_len);
307 opt_len -= l;
308 nd_opt = ((void *)nd_opt) + l;
310 return ndopts;
313 static inline u8 *ndisc_opt_addr_data(struct nd_opt_hdr *p,
314 struct net_device *dev)
316 u8 *lladdr = (u8 *)(p + 1);
317 int lladdrlen = p->nd_opt_len << 3;
318 int prepad = ndisc_addr_option_pad(dev->type);
319 if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len + prepad))
320 return NULL;
321 return (lladdr + prepad);
324 int ndisc_mc_map(struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
326 switch (dev->type) {
327 case ARPHRD_ETHER:
328 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
329 case ARPHRD_FDDI:
330 ipv6_eth_mc_map(addr, buf);
331 return 0;
332 case ARPHRD_IEEE802_TR:
333 ipv6_tr_mc_map(addr,buf);
334 return 0;
335 case ARPHRD_ARCNET:
336 ipv6_arcnet_mc_map(addr, buf);
337 return 0;
338 case ARPHRD_INFINIBAND:
339 ipv6_ib_mc_map(addr, dev->broadcast, buf);
340 return 0;
341 default:
342 if (dir) {
343 memcpy(buf, dev->broadcast, dev->addr_len);
344 return 0;
347 return -EINVAL;
350 EXPORT_SYMBOL(ndisc_mc_map);
352 static u32 ndisc_hash(const void *pkey, const struct net_device *dev)
354 const u32 *p32 = pkey;
355 u32 addr_hash, i;
357 addr_hash = 0;
358 for (i = 0; i < (sizeof(struct in6_addr) / sizeof(u32)); i++)
359 addr_hash ^= *p32++;
361 return jhash_2words(addr_hash, dev->ifindex, nd_tbl.hash_rnd);
364 static int ndisc_constructor(struct neighbour *neigh)
366 struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
367 struct net_device *dev = neigh->dev;
368 struct inet6_dev *in6_dev;
369 struct neigh_parms *parms;
370 int is_multicast = ipv6_addr_is_multicast(addr);
372 rcu_read_lock();
373 in6_dev = in6_dev_get(dev);
374 if (in6_dev == NULL) {
375 rcu_read_unlock();
376 return -EINVAL;
379 parms = in6_dev->nd_parms;
380 __neigh_parms_put(neigh->parms);
381 neigh->parms = neigh_parms_clone(parms);
382 rcu_read_unlock();
384 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
385 if (!dev->header_ops) {
386 neigh->nud_state = NUD_NOARP;
387 neigh->ops = &ndisc_direct_ops;
388 neigh->output = neigh->ops->queue_xmit;
389 } else {
390 if (is_multicast) {
391 neigh->nud_state = NUD_NOARP;
392 ndisc_mc_map(addr, neigh->ha, dev, 1);
393 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
394 neigh->nud_state = NUD_NOARP;
395 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
396 if (dev->flags&IFF_LOOPBACK)
397 neigh->type = RTN_LOCAL;
398 } else if (dev->flags&IFF_POINTOPOINT) {
399 neigh->nud_state = NUD_NOARP;
400 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
402 if (dev->header_ops->cache)
403 neigh->ops = &ndisc_hh_ops;
404 else
405 neigh->ops = &ndisc_generic_ops;
406 if (neigh->nud_state&NUD_VALID)
407 neigh->output = neigh->ops->connected_output;
408 else
409 neigh->output = neigh->ops->output;
411 in6_dev_put(in6_dev);
412 return 0;
415 static int pndisc_constructor(struct pneigh_entry *n)
417 struct in6_addr *addr = (struct in6_addr*)&n->key;
418 struct in6_addr maddr;
419 struct net_device *dev = n->dev;
421 if (dev == NULL || __in6_dev_get(dev) == NULL)
422 return -EINVAL;
423 addrconf_addr_solict_mult(addr, &maddr);
424 ipv6_dev_mc_inc(dev, &maddr);
425 return 0;
428 static void pndisc_destructor(struct pneigh_entry *n)
430 struct in6_addr *addr = (struct in6_addr*)&n->key;
431 struct in6_addr maddr;
432 struct net_device *dev = n->dev;
434 if (dev == NULL || __in6_dev_get(dev) == NULL)
435 return;
436 addrconf_addr_solict_mult(addr, &maddr);
437 ipv6_dev_mc_dec(dev, &maddr);
440 struct sk_buff *ndisc_build_skb(struct net_device *dev,
441 const struct in6_addr *daddr,
442 const struct in6_addr *saddr,
443 struct icmp6hdr *icmp6h,
444 const struct in6_addr *target,
445 int llinfo)
447 struct net *net = dev_net(dev);
448 struct sock *sk = net->ipv6.ndisc_sk;
449 struct sk_buff *skb;
450 struct icmp6hdr *hdr;
451 int len;
452 int err;
453 u8 *opt;
455 if (!dev->addr_len)
456 llinfo = 0;
458 len = sizeof(struct icmp6hdr) + (target ? sizeof(*target) : 0);
459 if (llinfo)
460 len += ndisc_opt_addr_space(dev);
462 skb = sock_alloc_send_skb(sk,
463 (MAX_HEADER + sizeof(struct ipv6hdr) +
464 len + LL_ALLOCATED_SPACE(dev)),
465 1, &err);
466 if (!skb) {
467 ND_PRINTK0(KERN_ERR
468 "ICMPv6 ND: %s() failed to allocate an skb, err=%d.\n",
469 __func__, err);
470 return NULL;
473 skb_reserve(skb, LL_RESERVED_SPACE(dev));
474 ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
476 skb->transport_header = skb->tail;
477 skb_put(skb, len);
479 hdr = (struct icmp6hdr *)skb_transport_header(skb);
480 memcpy(hdr, icmp6h, sizeof(*hdr));
482 opt = skb_transport_header(skb) + sizeof(struct icmp6hdr);
483 if (target) {
484 ipv6_addr_copy((struct in6_addr *)opt, target);
485 opt += sizeof(*target);
488 if (llinfo)
489 ndisc_fill_addr_option(opt, llinfo, dev->dev_addr,
490 dev->addr_len, dev->type);
492 hdr->icmp6_cksum = csum_ipv6_magic(saddr, daddr, len,
493 IPPROTO_ICMPV6,
494 csum_partial(hdr,
495 len, 0));
497 return skb;
500 EXPORT_SYMBOL(ndisc_build_skb);
502 void ndisc_send_skb(struct sk_buff *skb,
503 struct net_device *dev,
504 struct neighbour *neigh,
505 const struct in6_addr *daddr,
506 const struct in6_addr *saddr,
507 struct icmp6hdr *icmp6h)
509 struct flowi fl;
510 struct dst_entry *dst;
511 struct net *net = dev_net(dev);
512 struct sock *sk = net->ipv6.ndisc_sk;
513 struct inet6_dev *idev;
514 int err;
515 u8 type;
517 type = icmp6h->icmp6_type;
519 icmpv6_flow_init(sk, &fl, type, saddr, daddr, dev->ifindex);
521 dst = icmp6_dst_alloc(dev, neigh, daddr);
522 if (!dst) {
523 kfree_skb(skb);
524 return;
527 err = xfrm_lookup(net, &dst, &fl, NULL, 0);
528 if (err < 0) {
529 kfree_skb(skb);
530 return;
533 skb_dst_set(skb, dst);
535 idev = in6_dev_get(dst->dev);
536 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
538 err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, dst->dev,
539 dst_output);
540 if (!err) {
541 ICMP6MSGOUT_INC_STATS(net, idev, type);
542 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
545 if (likely(idev != NULL))
546 in6_dev_put(idev);
549 EXPORT_SYMBOL(ndisc_send_skb);
552 * Send a Neighbour Discover packet
554 static void __ndisc_send(struct net_device *dev,
555 struct neighbour *neigh,
556 const struct in6_addr *daddr,
557 const struct in6_addr *saddr,
558 struct icmp6hdr *icmp6h, const struct in6_addr *target,
559 int llinfo)
561 struct sk_buff *skb;
563 skb = ndisc_build_skb(dev, daddr, saddr, icmp6h, target, llinfo);
564 if (!skb)
565 return;
567 ndisc_send_skb(skb, dev, neigh, daddr, saddr, icmp6h);
570 static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
571 const struct in6_addr *daddr,
572 const struct in6_addr *solicited_addr,
573 int router, int solicited, int override, int inc_opt)
575 struct in6_addr tmpaddr;
576 struct inet6_ifaddr *ifp;
577 const struct in6_addr *src_addr;
578 struct icmp6hdr icmp6h = {
579 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
582 /* for anycast or proxy, solicited_addr != src_addr */
583 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
584 if (ifp) {
585 src_addr = solicited_addr;
586 if (ifp->flags & IFA_F_OPTIMISTIC)
587 override = 0;
588 in6_ifa_put(ifp);
589 } else {
590 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
591 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
592 &tmpaddr))
593 return;
594 src_addr = &tmpaddr;
597 icmp6h.icmp6_router = router;
598 icmp6h.icmp6_solicited = solicited;
599 icmp6h.icmp6_override = override;
601 __ndisc_send(dev, neigh, daddr, src_addr,
602 &icmp6h, solicited_addr,
603 inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
606 void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
607 const struct in6_addr *solicit,
608 const struct in6_addr *daddr, const struct in6_addr *saddr)
610 struct in6_addr addr_buf;
611 struct icmp6hdr icmp6h = {
612 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
615 if (saddr == NULL) {
616 if (ipv6_get_lladdr(dev, &addr_buf,
617 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
618 return;
619 saddr = &addr_buf;
622 __ndisc_send(dev, neigh, daddr, saddr,
623 &icmp6h, solicit,
624 !ipv6_addr_any(saddr) ? ND_OPT_SOURCE_LL_ADDR : 0);
627 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
628 const struct in6_addr *daddr)
630 struct icmp6hdr icmp6h = {
631 .icmp6_type = NDISC_ROUTER_SOLICITATION,
633 int send_sllao = dev->addr_len;
635 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
637 * According to section 2.2 of RFC 4429, we must not
638 * send router solicitations with a sllao from
639 * optimistic addresses, but we may send the solicitation
640 * if we don't include the sllao. So here we check
641 * if our address is optimistic, and if so, we
642 * suppress the inclusion of the sllao.
644 if (send_sllao) {
645 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
646 dev, 1);
647 if (ifp) {
648 if (ifp->flags & IFA_F_OPTIMISTIC) {
649 send_sllao = 0;
651 in6_ifa_put(ifp);
652 } else {
653 send_sllao = 0;
656 #endif
657 __ndisc_send(dev, NULL, daddr, saddr,
658 &icmp6h, NULL,
659 send_sllao ? ND_OPT_SOURCE_LL_ADDR : 0);
663 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
666 * "The sender MUST return an ICMP
667 * destination unreachable"
669 dst_link_failure(skb);
670 kfree_skb(skb);
673 /* Called with locked neigh: either read or both */
675 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
677 struct in6_addr *saddr = NULL;
678 struct in6_addr mcaddr;
679 struct net_device *dev = neigh->dev;
680 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
681 int probes = atomic_read(&neigh->probes);
683 if (skb && ipv6_chk_addr(dev_net(dev), &ipv6_hdr(skb)->saddr, dev, 1))
684 saddr = &ipv6_hdr(skb)->saddr;
686 if ((probes -= neigh->parms->ucast_probes) < 0) {
687 if (!(neigh->nud_state & NUD_VALID)) {
688 ND_PRINTK1(KERN_DEBUG "%s(): trying to ucast probe in NUD_INVALID: %pI6\n",
689 __func__, target);
691 ndisc_send_ns(dev, neigh, target, target, saddr);
692 } else if ((probes -= neigh->parms->app_probes) < 0) {
693 #ifdef CONFIG_ARPD
694 neigh_app_ns(neigh);
695 #endif
696 } else {
697 addrconf_addr_solict_mult(target, &mcaddr);
698 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
702 static int pndisc_is_router(const void *pkey,
703 struct net_device *dev)
705 struct pneigh_entry *n;
706 int ret = -1;
708 read_lock_bh(&nd_tbl.lock);
709 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
710 if (n)
711 ret = !!(n->flags & NTF_ROUTER);
712 read_unlock_bh(&nd_tbl.lock);
714 return ret;
717 static void ndisc_recv_ns(struct sk_buff *skb)
719 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
720 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
721 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
722 u8 *lladdr = NULL;
723 u32 ndoptlen = skb->tail - (skb->transport_header +
724 offsetof(struct nd_msg, opt));
725 struct ndisc_options ndopts;
726 struct net_device *dev = skb->dev;
727 struct inet6_ifaddr *ifp;
728 struct inet6_dev *idev = NULL;
729 struct neighbour *neigh;
730 int dad = ipv6_addr_any(saddr);
731 int inc;
732 int is_router = -1;
734 if (ipv6_addr_is_multicast(&msg->target)) {
735 ND_PRINTK2(KERN_WARNING
736 "ICMPv6 NS: multicast target address");
737 return;
741 * RFC2461 7.1.1:
742 * DAD has to be destined for solicited node multicast address.
744 if (dad &&
745 !(daddr->s6_addr32[0] == htonl(0xff020000) &&
746 daddr->s6_addr32[1] == htonl(0x00000000) &&
747 daddr->s6_addr32[2] == htonl(0x00000001) &&
748 daddr->s6_addr [12] == 0xff )) {
749 ND_PRINTK2(KERN_WARNING
750 "ICMPv6 NS: bad DAD packet (wrong destination)\n");
751 return;
754 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
755 ND_PRINTK2(KERN_WARNING
756 "ICMPv6 NS: invalid ND options\n");
757 return;
760 if (ndopts.nd_opts_src_lladdr) {
761 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
762 if (!lladdr) {
763 ND_PRINTK2(KERN_WARNING
764 "ICMPv6 NS: invalid link-layer address length\n");
765 return;
768 /* RFC2461 7.1.1:
769 * If the IP source address is the unspecified address,
770 * there MUST NOT be source link-layer address option
771 * in the message.
773 if (dad) {
774 ND_PRINTK2(KERN_WARNING
775 "ICMPv6 NS: bad DAD packet (link-layer address option)\n");
776 return;
780 inc = ipv6_addr_is_multicast(daddr);
782 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
783 if (ifp) {
785 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
786 if (dad) {
787 if (dev->type == ARPHRD_IEEE802_TR) {
788 const unsigned char *sadr;
789 sadr = skb_mac_header(skb);
790 if (((sadr[8] ^ dev->dev_addr[0]) & 0x7f) == 0 &&
791 sadr[9] == dev->dev_addr[1] &&
792 sadr[10] == dev->dev_addr[2] &&
793 sadr[11] == dev->dev_addr[3] &&
794 sadr[12] == dev->dev_addr[4] &&
795 sadr[13] == dev->dev_addr[5]) {
796 /* looped-back to us */
797 goto out;
802 * We are colliding with another node
803 * who is doing DAD
804 * so fail our DAD process
806 addrconf_dad_failure(ifp);
807 return;
808 } else {
810 * This is not a dad solicitation.
811 * If we are an optimistic node,
812 * we should respond.
813 * Otherwise, we should ignore it.
815 if (!(ifp->flags & IFA_F_OPTIMISTIC))
816 goto out;
820 idev = ifp->idev;
821 } else {
822 struct net *net = dev_net(dev);
824 idev = in6_dev_get(dev);
825 if (!idev) {
826 /* XXX: count this drop? */
827 return;
830 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
831 (idev->cnf.forwarding &&
832 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
833 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
834 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
835 skb->pkt_type != PACKET_HOST &&
836 inc != 0 &&
837 idev->nd_parms->proxy_delay != 0) {
839 * for anycast or proxy,
840 * sender should delay its response
841 * by a random time between 0 and
842 * MAX_ANYCAST_DELAY_TIME seconds.
843 * (RFC2461) -- yoshfuji
845 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
846 if (n)
847 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
848 goto out;
850 } else
851 goto out;
854 if (is_router < 0)
855 is_router = !!idev->cnf.forwarding;
857 if (dad) {
858 ndisc_send_na(dev, NULL, &in6addr_linklocal_allnodes, &msg->target,
859 is_router, 0, (ifp != NULL), 1);
860 goto out;
863 if (inc)
864 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
865 else
866 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
869 * update / create cache entry
870 * for the source address
872 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
873 !inc || lladdr || !dev->addr_len);
874 if (neigh)
875 neigh_update(neigh, lladdr, NUD_STALE,
876 NEIGH_UPDATE_F_WEAK_OVERRIDE|
877 NEIGH_UPDATE_F_OVERRIDE);
878 if (neigh || !dev->header_ops) {
879 ndisc_send_na(dev, neigh, saddr, &msg->target,
880 is_router,
881 1, (ifp != NULL && inc), inc);
882 if (neigh)
883 neigh_release(neigh);
886 out:
887 if (ifp)
888 in6_ifa_put(ifp);
889 else
890 in6_dev_put(idev);
892 return;
895 static void ndisc_recv_na(struct sk_buff *skb)
897 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
898 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
899 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
900 u8 *lladdr = NULL;
901 u32 ndoptlen = skb->tail - (skb->transport_header +
902 offsetof(struct nd_msg, opt));
903 struct ndisc_options ndopts;
904 struct net_device *dev = skb->dev;
905 struct inet6_ifaddr *ifp;
906 struct neighbour *neigh;
908 if (skb->len < sizeof(struct nd_msg)) {
909 ND_PRINTK2(KERN_WARNING
910 "ICMPv6 NA: packet too short\n");
911 return;
914 if (ipv6_addr_is_multicast(&msg->target)) {
915 ND_PRINTK2(KERN_WARNING
916 "ICMPv6 NA: target address is multicast.\n");
917 return;
920 if (ipv6_addr_is_multicast(daddr) &&
921 msg->icmph.icmp6_solicited) {
922 ND_PRINTK2(KERN_WARNING
923 "ICMPv6 NA: solicited NA is multicasted.\n");
924 return;
927 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
928 ND_PRINTK2(KERN_WARNING
929 "ICMPv6 NS: invalid ND option\n");
930 return;
932 if (ndopts.nd_opts_tgt_lladdr) {
933 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
934 if (!lladdr) {
935 ND_PRINTK2(KERN_WARNING
936 "ICMPv6 NA: invalid link-layer address length\n");
937 return;
940 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
941 if (ifp) {
942 if (ifp->flags & IFA_F_TENTATIVE) {
943 addrconf_dad_failure(ifp);
944 return;
946 /* What should we make now? The advertisement
947 is invalid, but ndisc specs say nothing
948 about it. It could be misconfiguration, or
949 an smart proxy agent tries to help us :-)
951 We should not print the error if NA has been
952 received from loopback - it is just our own
953 unsolicited advertisement.
955 if (skb->pkt_type != PACKET_LOOPBACK)
956 ND_PRINTK1(KERN_WARNING
957 "ICMPv6 NA: someone advertises our address %pI6 on %s!\n",
958 &ifp->addr, ifp->idev->dev->name);
959 in6_ifa_put(ifp);
960 return;
962 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
964 if (neigh) {
965 u8 old_flags = neigh->flags;
966 struct net *net = dev_net(dev);
968 if (neigh->nud_state & NUD_FAILED)
969 goto out;
972 * Don't update the neighbor cache entry on a proxy NA from
973 * ourselves because either the proxied node is off link or it
974 * has already sent a NA to us.
976 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
977 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
978 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
979 /* XXX: idev->cnf.prixy_ndp */
980 goto out;
983 neigh_update(neigh, lladdr,
984 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
985 NEIGH_UPDATE_F_WEAK_OVERRIDE|
986 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
987 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
988 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
990 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
992 * Change: router to host
994 struct rt6_info *rt;
995 rt = rt6_get_dflt_router(saddr, dev);
996 if (rt)
997 ip6_del_rt(rt);
1000 out:
1001 neigh_release(neigh);
1005 static void ndisc_recv_rs(struct sk_buff *skb)
1007 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1008 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1009 struct neighbour *neigh;
1010 struct inet6_dev *idev;
1011 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1012 struct ndisc_options ndopts;
1013 u8 *lladdr = NULL;
1015 if (skb->len < sizeof(*rs_msg))
1016 return;
1018 idev = in6_dev_get(skb->dev);
1019 if (!idev) {
1020 if (net_ratelimit())
1021 ND_PRINTK1("ICMP6 RS: can't find in6 device\n");
1022 return;
1025 /* Don't accept RS if we're not in router mode */
1026 if (!idev->cnf.forwarding)
1027 goto out;
1030 * Don't update NCE if src = ::;
1031 * this implies that the source node has no ip address assigned yet.
1033 if (ipv6_addr_any(saddr))
1034 goto out;
1036 /* Parse ND options */
1037 if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
1038 if (net_ratelimit())
1039 ND_PRINTK2("ICMP6 NS: invalid ND option, ignored\n");
1040 goto out;
1043 if (ndopts.nd_opts_src_lladdr) {
1044 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1045 skb->dev);
1046 if (!lladdr)
1047 goto out;
1050 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1051 if (neigh) {
1052 neigh_update(neigh, lladdr, NUD_STALE,
1053 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1054 NEIGH_UPDATE_F_OVERRIDE|
1055 NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1056 neigh_release(neigh);
1058 out:
1059 in6_dev_put(idev);
1062 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1064 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1065 struct sk_buff *skb;
1066 struct nlmsghdr *nlh;
1067 struct nduseroptmsg *ndmsg;
1068 struct net *net = dev_net(ra->dev);
1069 int err;
1070 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1071 + (opt->nd_opt_len << 3));
1072 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1074 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1075 if (skb == NULL) {
1076 err = -ENOBUFS;
1077 goto errout;
1080 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1081 if (nlh == NULL) {
1082 goto nla_put_failure;
1085 ndmsg = nlmsg_data(nlh);
1086 ndmsg->nduseropt_family = AF_INET6;
1087 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1088 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1089 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1090 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1092 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1094 NLA_PUT(skb, NDUSEROPT_SRCADDR, sizeof(struct in6_addr),
1095 &ipv6_hdr(ra)->saddr);
1096 nlmsg_end(skb, nlh);
1098 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1099 return;
1101 nla_put_failure:
1102 nlmsg_free(skb);
1103 err = -EMSGSIZE;
1104 errout:
1105 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1108 static void ndisc_router_discovery(struct sk_buff *skb)
1110 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1111 struct neighbour *neigh = NULL;
1112 struct inet6_dev *in6_dev;
1113 struct rt6_info *rt = NULL;
1114 int lifetime;
1115 struct ndisc_options ndopts;
1116 int optlen;
1117 unsigned int pref = 0;
1119 __u8 * opt = (__u8 *)(ra_msg + 1);
1121 optlen = (skb->tail - skb->transport_header) - sizeof(struct ra_msg);
1123 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1124 ND_PRINTK2(KERN_WARNING
1125 "ICMPv6 RA: source address is not link-local.\n");
1126 return;
1128 if (optlen < 0) {
1129 ND_PRINTK2(KERN_WARNING
1130 "ICMPv6 RA: packet too short\n");
1131 return;
1134 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1135 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1136 ND_PRINTK2(KERN_WARNING
1137 "ICMPv6 RA: from host or unauthorized router\n");
1138 return;
1140 #endif
1143 * set the RA_RECV flag in the interface
1146 in6_dev = in6_dev_get(skb->dev);
1147 if (in6_dev == NULL) {
1148 ND_PRINTK0(KERN_ERR
1149 "ICMPv6 RA: can't find inet6 device for %s.\n",
1150 skb->dev->name);
1151 return;
1154 if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1155 in6_dev_put(in6_dev);
1156 ND_PRINTK2(KERN_WARNING
1157 "ICMP6 RA: invalid ND options\n");
1158 return;
1161 /* skip route and link configuration on routers */
1162 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_ra)
1163 goto skip_linkparms;
1165 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1166 /* skip link-specific parameters from interior routers */
1167 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1168 goto skip_linkparms;
1169 #endif
1171 if (in6_dev->if_flags & IF_RS_SENT) {
1173 * flag that an RA was received after an RS was sent
1174 * out on this interface.
1176 in6_dev->if_flags |= IF_RA_RCVD;
1180 * Remember the managed/otherconf flags from most recently
1181 * received RA message (RFC 2462) -- yoshfuji
1183 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1184 IF_RA_OTHERCONF)) |
1185 (ra_msg->icmph.icmp6_addrconf_managed ?
1186 IF_RA_MANAGED : 0) |
1187 (ra_msg->icmph.icmp6_addrconf_other ?
1188 IF_RA_OTHERCONF : 0);
1190 if (!in6_dev->cnf.accept_ra_defrtr)
1191 goto skip_defrtr;
1193 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1195 #ifdef CONFIG_IPV6_ROUTER_PREF
1196 pref = ra_msg->icmph.icmp6_router_pref;
1197 /* 10b is handled as if it were 00b (medium) */
1198 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1199 !in6_dev->cnf.accept_ra_rtr_pref)
1200 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1201 #endif
1203 rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1205 if (rt)
1206 neigh = rt->rt6i_nexthop;
1208 if (rt && lifetime == 0) {
1209 neigh_clone(neigh);
1210 ip6_del_rt(rt);
1211 rt = NULL;
1214 if (rt == NULL && lifetime) {
1215 ND_PRINTK3(KERN_DEBUG
1216 "ICMPv6 RA: adding default router.\n");
1218 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1219 if (rt == NULL) {
1220 ND_PRINTK0(KERN_ERR
1221 "ICMPv6 RA: %s() failed to add default route.\n",
1222 __func__);
1223 in6_dev_put(in6_dev);
1224 return;
1227 neigh = rt->rt6i_nexthop;
1228 if (neigh == NULL) {
1229 ND_PRINTK0(KERN_ERR
1230 "ICMPv6 RA: %s() got default router without neighbour.\n",
1231 __func__);
1232 dst_release(&rt->u.dst);
1233 in6_dev_put(in6_dev);
1234 return;
1236 neigh->flags |= NTF_ROUTER;
1237 } else if (rt) {
1238 rt->rt6i_flags = (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1241 if (rt)
1242 rt->rt6i_expires = jiffies + (HZ * lifetime);
1244 if (ra_msg->icmph.icmp6_hop_limit) {
1245 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1246 if (rt)
1247 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = ra_msg->icmph.icmp6_hop_limit;
1250 skip_defrtr:
1253 * Update Reachable Time and Retrans Timer
1256 if (in6_dev->nd_parms) {
1257 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1259 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1260 rtime = (rtime*HZ)/1000;
1261 if (rtime < HZ/10)
1262 rtime = HZ/10;
1263 in6_dev->nd_parms->retrans_time = rtime;
1264 in6_dev->tstamp = jiffies;
1265 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1268 rtime = ntohl(ra_msg->reachable_time);
1269 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1270 rtime = (rtime*HZ)/1000;
1272 if (rtime < HZ/10)
1273 rtime = HZ/10;
1275 if (rtime != in6_dev->nd_parms->base_reachable_time) {
1276 in6_dev->nd_parms->base_reachable_time = rtime;
1277 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1278 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1279 in6_dev->tstamp = jiffies;
1280 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1285 skip_linkparms:
1288 * Process options.
1291 if (!neigh)
1292 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1293 skb->dev, 1);
1294 if (neigh) {
1295 u8 *lladdr = NULL;
1296 if (ndopts.nd_opts_src_lladdr) {
1297 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1298 skb->dev);
1299 if (!lladdr) {
1300 ND_PRINTK2(KERN_WARNING
1301 "ICMPv6 RA: invalid link-layer address length\n");
1302 goto out;
1305 neigh_update(neigh, lladdr, NUD_STALE,
1306 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1307 NEIGH_UPDATE_F_OVERRIDE|
1308 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1309 NEIGH_UPDATE_F_ISROUTER);
1312 /* skip route and link configuration on routers */
1313 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_ra)
1314 goto out;
1316 #ifdef CONFIG_IPV6_ROUTE_INFO
1317 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1318 struct nd_opt_hdr *p;
1319 for (p = ndopts.nd_opts_ri;
1321 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1322 struct route_info *ri = (struct route_info *)p;
1323 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1324 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1325 ri->prefix_len == 0)
1326 continue;
1327 #endif
1328 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1329 continue;
1330 rt6_route_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3,
1331 &ipv6_hdr(skb)->saddr);
1334 #endif
1336 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1337 /* skip link-specific ndopts from interior routers */
1338 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1339 goto out;
1340 #endif
1342 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1343 struct nd_opt_hdr *p;
1344 for (p = ndopts.nd_opts_pi;
1346 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1347 addrconf_prefix_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3);
1351 if (ndopts.nd_opts_mtu) {
1352 __be32 n;
1353 u32 mtu;
1355 memcpy(&n, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1356 mtu = ntohl(n);
1358 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1359 ND_PRINTK2(KERN_WARNING
1360 "ICMPv6 RA: invalid mtu: %d\n",
1361 mtu);
1362 } else if (in6_dev->cnf.mtu6 != mtu) {
1363 in6_dev->cnf.mtu6 = mtu;
1365 if (rt)
1366 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
1368 rt6_mtu_change(skb->dev, mtu);
1372 if (ndopts.nd_useropts) {
1373 struct nd_opt_hdr *p;
1374 for (p = ndopts.nd_useropts;
1376 p = ndisc_next_useropt(p, ndopts.nd_useropts_end)) {
1377 ndisc_ra_useropt(skb, p);
1381 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1382 ND_PRINTK2(KERN_WARNING
1383 "ICMPv6 RA: invalid RA options");
1385 out:
1386 if (rt)
1387 dst_release(&rt->u.dst);
1388 else if (neigh)
1389 neigh_release(neigh);
1390 in6_dev_put(in6_dev);
1393 static void ndisc_redirect_rcv(struct sk_buff *skb)
1395 struct inet6_dev *in6_dev;
1396 struct icmp6hdr *icmph;
1397 struct in6_addr *dest;
1398 struct in6_addr *target; /* new first hop to destination */
1399 struct neighbour *neigh;
1400 int on_link = 0;
1401 struct ndisc_options ndopts;
1402 int optlen;
1403 u8 *lladdr = NULL;
1405 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1406 switch (skb->ndisc_nodetype) {
1407 case NDISC_NODETYPE_HOST:
1408 case NDISC_NODETYPE_NODEFAULT:
1409 ND_PRINTK2(KERN_WARNING
1410 "ICMPv6 Redirect: from host or unauthorized router\n");
1411 return;
1413 #endif
1415 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1416 ND_PRINTK2(KERN_WARNING
1417 "ICMPv6 Redirect: source address is not link-local.\n");
1418 return;
1421 optlen = skb->tail - skb->transport_header;
1422 optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1424 if (optlen < 0) {
1425 ND_PRINTK2(KERN_WARNING
1426 "ICMPv6 Redirect: packet too short\n");
1427 return;
1430 icmph = icmp6_hdr(skb);
1431 target = (struct in6_addr *) (icmph + 1);
1432 dest = target + 1;
1434 if (ipv6_addr_is_multicast(dest)) {
1435 ND_PRINTK2(KERN_WARNING
1436 "ICMPv6 Redirect: destination address is multicast.\n");
1437 return;
1440 if (ipv6_addr_equal(dest, target)) {
1441 on_link = 1;
1442 } else if (ipv6_addr_type(target) !=
1443 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1444 ND_PRINTK2(KERN_WARNING
1445 "ICMPv6 Redirect: target address is not link-local unicast.\n");
1446 return;
1449 in6_dev = in6_dev_get(skb->dev);
1450 if (!in6_dev)
1451 return;
1452 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects) {
1453 in6_dev_put(in6_dev);
1454 return;
1457 /* RFC2461 8.1:
1458 * The IP source address of the Redirect MUST be the same as the current
1459 * first-hop router for the specified ICMP Destination Address.
1462 if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1463 ND_PRINTK2(KERN_WARNING
1464 "ICMPv6 Redirect: invalid ND options\n");
1465 in6_dev_put(in6_dev);
1466 return;
1468 if (ndopts.nd_opts_tgt_lladdr) {
1469 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1470 skb->dev);
1471 if (!lladdr) {
1472 ND_PRINTK2(KERN_WARNING
1473 "ICMPv6 Redirect: invalid link-layer address length\n");
1474 in6_dev_put(in6_dev);
1475 return;
1479 neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1480 if (neigh) {
1481 rt6_redirect(dest, &ipv6_hdr(skb)->daddr,
1482 &ipv6_hdr(skb)->saddr, neigh, lladdr,
1483 on_link);
1484 neigh_release(neigh);
1486 in6_dev_put(in6_dev);
1489 void ndisc_send_redirect(struct sk_buff *skb, struct neighbour *neigh,
1490 const struct in6_addr *target)
1492 struct net_device *dev = skb->dev;
1493 struct net *net = dev_net(dev);
1494 struct sock *sk = net->ipv6.ndisc_sk;
1495 int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1496 struct sk_buff *buff;
1497 struct icmp6hdr *icmph;
1498 struct in6_addr saddr_buf;
1499 struct in6_addr *addrp;
1500 struct rt6_info *rt;
1501 struct dst_entry *dst;
1502 struct inet6_dev *idev;
1503 struct flowi fl;
1504 u8 *opt;
1505 int rd_len;
1506 int err;
1507 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1509 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1510 ND_PRINTK2(KERN_WARNING
1511 "ICMPv6 Redirect: no link-local address on %s\n",
1512 dev->name);
1513 return;
1516 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1517 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1518 ND_PRINTK2(KERN_WARNING
1519 "ICMPv6 Redirect: target address is not link-local unicast.\n");
1520 return;
1523 icmpv6_flow_init(sk, &fl, NDISC_REDIRECT,
1524 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1526 dst = ip6_route_output(net, NULL, &fl);
1527 if (dst == NULL)
1528 return;
1530 err = xfrm_lookup(net, &dst, &fl, NULL, 0);
1531 if (err)
1532 return;
1534 rt = (struct rt6_info *) dst;
1536 if (rt->rt6i_flags & RTF_GATEWAY) {
1537 ND_PRINTK2(KERN_WARNING
1538 "ICMPv6 Redirect: destination is not a neighbour.\n");
1539 goto release;
1541 if (!xrlim_allow(dst, 1*HZ))
1542 goto release;
1544 if (dev->addr_len) {
1545 read_lock_bh(&neigh->lock);
1546 if (neigh->nud_state & NUD_VALID) {
1547 memcpy(ha_buf, neigh->ha, dev->addr_len);
1548 read_unlock_bh(&neigh->lock);
1549 ha = ha_buf;
1550 len += ndisc_opt_addr_space(dev);
1551 } else
1552 read_unlock_bh(&neigh->lock);
1555 rd_len = min_t(unsigned int,
1556 IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1557 rd_len &= ~0x7;
1558 len += rd_len;
1560 buff = sock_alloc_send_skb(sk,
1561 (MAX_HEADER + sizeof(struct ipv6hdr) +
1562 len + LL_ALLOCATED_SPACE(dev)),
1563 1, &err);
1564 if (buff == NULL) {
1565 ND_PRINTK0(KERN_ERR
1566 "ICMPv6 Redirect: %s() failed to allocate an skb, err=%d.\n",
1567 __func__, err);
1568 goto release;
1571 skb_reserve(buff, LL_RESERVED_SPACE(dev));
1572 ip6_nd_hdr(sk, buff, dev, &saddr_buf, &ipv6_hdr(skb)->saddr,
1573 IPPROTO_ICMPV6, len);
1575 skb_set_transport_header(buff, skb_tail_pointer(buff) - buff->data);
1576 skb_put(buff, len);
1577 icmph = icmp6_hdr(buff);
1579 memset(icmph, 0, sizeof(struct icmp6hdr));
1580 icmph->icmp6_type = NDISC_REDIRECT;
1583 * copy target and destination addresses
1586 addrp = (struct in6_addr *)(icmph + 1);
1587 ipv6_addr_copy(addrp, target);
1588 addrp++;
1589 ipv6_addr_copy(addrp, &ipv6_hdr(skb)->daddr);
1591 opt = (u8*) (addrp + 1);
1594 * include target_address option
1597 if (ha)
1598 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1599 dev->addr_len, dev->type);
1602 * build redirect option and copy skb over to the new packet.
1605 memset(opt, 0, 8);
1606 *(opt++) = ND_OPT_REDIRECT_HDR;
1607 *(opt++) = (rd_len >> 3);
1608 opt += 6;
1610 memcpy(opt, ipv6_hdr(skb), rd_len - 8);
1612 icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &ipv6_hdr(skb)->saddr,
1613 len, IPPROTO_ICMPV6,
1614 csum_partial(icmph, len, 0));
1616 skb_dst_set(buff, dst);
1617 idev = in6_dev_get(dst->dev);
1618 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1619 err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, buff, NULL, dst->dev,
1620 dst_output);
1621 if (!err) {
1622 ICMP6MSGOUT_INC_STATS(net, idev, NDISC_REDIRECT);
1623 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1626 if (likely(idev != NULL))
1627 in6_dev_put(idev);
1628 return;
1630 release:
1631 dst_release(dst);
1634 static void pndisc_redo(struct sk_buff *skb)
1636 ndisc_recv_ns(skb);
1637 kfree_skb(skb);
1640 int ndisc_rcv(struct sk_buff *skb)
1642 struct nd_msg *msg;
1644 if (!pskb_may_pull(skb, skb->len))
1645 return 0;
1647 msg = (struct nd_msg *)skb_transport_header(skb);
1649 __skb_push(skb, skb->data - skb_transport_header(skb));
1651 if (ipv6_hdr(skb)->hop_limit != 255) {
1652 ND_PRINTK2(KERN_WARNING
1653 "ICMPv6 NDISC: invalid hop-limit: %d\n",
1654 ipv6_hdr(skb)->hop_limit);
1655 return 0;
1658 if (msg->icmph.icmp6_code != 0) {
1659 ND_PRINTK2(KERN_WARNING
1660 "ICMPv6 NDISC: invalid ICMPv6 code: %d\n",
1661 msg->icmph.icmp6_code);
1662 return 0;
1665 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1667 switch (msg->icmph.icmp6_type) {
1668 case NDISC_NEIGHBOUR_SOLICITATION:
1669 ndisc_recv_ns(skb);
1670 break;
1672 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1673 ndisc_recv_na(skb);
1674 break;
1676 case NDISC_ROUTER_SOLICITATION:
1677 ndisc_recv_rs(skb);
1678 break;
1680 case NDISC_ROUTER_ADVERTISEMENT:
1681 ndisc_router_discovery(skb);
1682 break;
1684 case NDISC_REDIRECT:
1685 ndisc_redirect_rcv(skb);
1686 break;
1689 return 0;
1692 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1694 struct net_device *dev = ptr;
1695 struct net *net = dev_net(dev);
1697 switch (event) {
1698 case NETDEV_CHANGEADDR:
1699 neigh_changeaddr(&nd_tbl, dev);
1700 fib6_run_gc(~0UL, net);
1701 break;
1702 case NETDEV_DOWN:
1703 neigh_ifdown(&nd_tbl, dev);
1704 fib6_run_gc(~0UL, net);
1705 break;
1706 default:
1707 break;
1710 return NOTIFY_DONE;
1713 static struct notifier_block ndisc_netdev_notifier = {
1714 .notifier_call = ndisc_netdev_event,
1717 #ifdef CONFIG_SYSCTL
1718 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1719 const char *func, const char *dev_name)
1721 static char warncomm[TASK_COMM_LEN];
1722 static int warned;
1723 if (strcmp(warncomm, current->comm) && warned < 5) {
1724 strcpy(warncomm, current->comm);
1725 printk(KERN_WARNING
1726 "process `%s' is using deprecated sysctl (%s) "
1727 "net.ipv6.neigh.%s.%s; "
1728 "Use net.ipv6.neigh.%s.%s_ms "
1729 "instead.\n",
1730 warncomm, func,
1731 dev_name, ctl->procname,
1732 dev_name, ctl->procname);
1733 warned++;
1737 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1739 struct net_device *dev = ctl->extra1;
1740 struct inet6_dev *idev;
1741 int ret;
1743 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1744 (strcmp(ctl->procname, "base_reachable_time") == 0))
1745 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1747 if (strcmp(ctl->procname, "retrans_time") == 0)
1748 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1750 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1751 ret = proc_dointvec_jiffies(ctl, write,
1752 buffer, lenp, ppos);
1754 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1755 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1756 ret = proc_dointvec_ms_jiffies(ctl, write,
1757 buffer, lenp, ppos);
1758 else
1759 ret = -1;
1761 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1762 if (ctl->data == &idev->nd_parms->base_reachable_time)
1763 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1764 idev->tstamp = jiffies;
1765 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1766 in6_dev_put(idev);
1768 return ret;
1771 int ndisc_ifinfo_sysctl_strategy(ctl_table *ctl,
1772 void __user *oldval, size_t __user *oldlenp,
1773 void __user *newval, size_t newlen)
1775 struct net_device *dev = ctl->extra1;
1776 struct inet6_dev *idev;
1777 int ret;
1779 if (ctl->ctl_name == NET_NEIGH_RETRANS_TIME ||
1780 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME)
1781 ndisc_warn_deprecated_sysctl(ctl, "procfs", dev ? dev->name : "default");
1783 switch (ctl->ctl_name) {
1784 case NET_NEIGH_REACHABLE_TIME:
1785 ret = sysctl_jiffies(ctl, oldval, oldlenp, newval, newlen);
1786 break;
1787 case NET_NEIGH_RETRANS_TIME_MS:
1788 case NET_NEIGH_REACHABLE_TIME_MS:
1789 ret = sysctl_ms_jiffies(ctl, oldval, oldlenp, newval, newlen);
1790 break;
1791 default:
1792 ret = 0;
1795 if (newval && newlen && ret > 0 &&
1796 dev && (idev = in6_dev_get(dev)) != NULL) {
1797 if (ctl->ctl_name == NET_NEIGH_REACHABLE_TIME ||
1798 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME_MS)
1799 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1800 idev->tstamp = jiffies;
1801 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1802 in6_dev_put(idev);
1805 return ret;
1808 #endif
1810 static int ndisc_net_init(struct net *net)
1812 struct ipv6_pinfo *np;
1813 struct sock *sk;
1814 int err;
1816 err = inet_ctl_sock_create(&sk, PF_INET6,
1817 SOCK_RAW, IPPROTO_ICMPV6, net);
1818 if (err < 0) {
1819 ND_PRINTK0(KERN_ERR
1820 "ICMPv6 NDISC: Failed to initialize the control socket (err %d).\n",
1821 err);
1822 return err;
1825 net->ipv6.ndisc_sk = sk;
1827 np = inet6_sk(sk);
1828 np->hop_limit = 255;
1829 /* Do not loopback ndisc messages */
1830 np->mc_loop = 0;
1832 return 0;
1835 static void ndisc_net_exit(struct net *net)
1837 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1840 static struct pernet_operations ndisc_net_ops = {
1841 .init = ndisc_net_init,
1842 .exit = ndisc_net_exit,
1845 int __init ndisc_init(void)
1847 int err;
1849 err = register_pernet_subsys(&ndisc_net_ops);
1850 if (err)
1851 return err;
1853 * Initialize the neighbour table
1855 neigh_table_init(&nd_tbl);
1857 #ifdef CONFIG_SYSCTL
1858 err = neigh_sysctl_register(NULL, &nd_tbl.parms, NET_IPV6,
1859 NET_IPV6_NEIGH, "ipv6",
1860 &ndisc_ifinfo_sysctl_change,
1861 &ndisc_ifinfo_sysctl_strategy);
1862 if (err)
1863 goto out_unregister_pernet;
1864 #endif
1865 err = register_netdevice_notifier(&ndisc_netdev_notifier);
1866 if (err)
1867 goto out_unregister_sysctl;
1868 out:
1869 return err;
1871 out_unregister_sysctl:
1872 #ifdef CONFIG_SYSCTL
1873 neigh_sysctl_unregister(&nd_tbl.parms);
1874 out_unregister_pernet:
1875 #endif
1876 unregister_pernet_subsys(&ndisc_net_ops);
1877 goto out;
1880 void ndisc_cleanup(void)
1882 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1883 #ifdef CONFIG_SYSCTL
1884 neigh_sysctl_unregister(&nd_tbl.parms);
1885 #endif
1886 neigh_table_clear(&nd_tbl);
1887 unregister_pernet_subsys(&ndisc_net_ops);